13.07.2015 Views

In-flight upset - 154 km west of Learmonth, WA, 7 October 2008,

In-flight upset - 154 km west of Learmonth, WA, 7 October 2008,

In-flight upset - 154 km west of Learmonth, WA, 7 October 2008,

SHOW MORE
SHOW LESS
  • No tags were found...

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

To conduct an effective fault tree analysis or FMEA requires safety analysts to havea very good understanding <strong>of</strong> the system and all its components. Experts have notedthat design engineers and safety analysts generally work with different views <strong>of</strong> thesystem’s design. This can lead to potential problems with consistency andcompleteness <strong>of</strong> understanding between the two groups. A recent US NationalAeronautics and Space Administration (NASA) research report (Joshi et al. 2006)noted:Safety engineers traditionally perform analysis, such as fault tree analysis,based on information synthesized from several sources, including informaldesign models and requirements documents. Unfortunately, these analyses arehighly subjective and dependent on the skill <strong>of</strong> the engineer. Fault trees areone <strong>of</strong> the most common techniques used by safety engineers, yet differentsafety engineers will <strong>of</strong>ten produce fault trees for the same system that differin substantive ways. The final fault tree is <strong>of</strong>ten produced only through aprocess <strong>of</strong> review and consensus building between the system and safetyengineers. Even after a consensus is reached, it is unlikely that the analysisresults will be complete, consistent, and error free due in part to the informalmodels used as the basis <strong>of</strong> the analysis. <strong>In</strong> fact, the lack <strong>of</strong> precise models <strong>of</strong>the system architecture and its failure modes <strong>of</strong>ten forces the safety analysts todevote much <strong>of</strong> their effort to gathering information about the systemarchitecture and system behavior and embedding this information in the safetyartifacts such as the fault trees.<strong>In</strong> summary, the task <strong>of</strong> identifying scenarios that lead to failure conditions requiresexpertise and a detailed knowledge <strong>of</strong> the proposed system design. However, assystem designs get more complex, then traditional safety assessment activitiesbecome inefficient and difficult to use successfully.Formal methods and model-based safety analysisThe term ‘formal methods’ refers to the use <strong>of</strong> mathematical techniques in thedesign and evaluation <strong>of</strong> complex systems. NASA (1995) stated:Formal Methods (FM) consist <strong>of</strong> a set <strong>of</strong> techniques and tools based onmathematical modeling and formal logic that are used to specify and verifyrequirements and designs for computer systems and s<strong>of</strong>tware. The use <strong>of</strong> FMon a project can assume various forms, ranging from occasional mathematicalnotation embedded in English specifications, to fully formal specificationsusing specification languages with a precise semantics. At their most rigorous,FM involve computer-assisted pro<strong>of</strong>s <strong>of</strong> key properties regarding the behavior<strong>of</strong> the system.The use <strong>of</strong> formal methods in the development <strong>of</strong> complex systems has beengrowing for over 20 years. For aviation applications, the use <strong>of</strong> formal methods wasinitially focused on the development <strong>of</strong> requirements. For example, thedevelopment <strong>of</strong> the functional requirements for the A320 and A330/A340 werebased on formal methods, and these were used to develop models that could be usedto conduct simulations for validating the EFCS system design (section 2.4.3).<strong>In</strong> recent years, there has been considerable interest in using formal methods tocreate a system model, and then using the model to automate some <strong>of</strong> the safetyassessment tasks. Proponents have argued that this approach enables the designengineers and safety analysts to work with the same view <strong>of</strong> the system, and toprovide better assurance that design errors will be identified in a more efficientmanner (Pumphrey 2001, Joshi et al. 2006).- 105 -

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!